Adaptive Description of ATM Traffic Flows

Abstract


 
 
 Today, the great challenge in broad band networks, such as ATM, is to support applications that need guaranteed QoS. Efficient resource allocation for connections that need guaranteed QoS is only possible if the application can provide a precise and reliable traffic descriptor that the network can understand. This work presents a methodology to obtain the parameters of the traffic descriptor of an ATM application from an initial sample of the flow. Based on real-time measurements of the flow, we suggest periodic renegotiations of the parameters, every time they vary above or below their predefined limits. We apply this method to a VBR MPEG video flow and demonstrate its performance. Observed results are discussed and some future uses of the methodology are suggested.